For example, Japanese Patent Gazette No. 2778374 discloses a device configured to generate data of a new road, which is not included in road map data, based on a travel locus of a vehicle. The device is further configured to store the generated data of the new road. Detection of a travel locus of a vehicle certainly causes an error. Therefore, when data of a new road is generated based on a travel locus of a vehicle, as disclosed in Japanese Patent Gazette No. 2778374, the generated data of the new road is not necessarily correct. Therefore, it is conceived that a user may confirm whether the generated data of the new road is erroneous, and the user may delete the generated data of the new road when determining the generated data to be erroneous, as a result of the confirmation.
However, even when the user deletes the generated data of the new road, erroneous learning may be performed again to generate data. In this case, generate data of the new road caused by the erroneous learning needs to be further deleted. Consequently, the user may be annoyed by repeatedly generated unnecessary data. In addition, since erroneous data of the new road exists, a correct new road, which is similar to the erroneously detected new road, may be determined to be already learned. In this case, learning of the correct new road may not be performed.